Joon Sung Park

Joon Sung Park is a computer scientist who builds simulations of societies.¹ He uses them to test the assumptions about human behaviour on which institutions and policies rest.¹ His research introduced generative agents: actors built on language models that keep a memory of what they observe and plan their days.¹ He also developed ways to ground them in real data and to check their simulated attitudes and behaviour against what real people did.¹ He runs population-scale simulations with agents representative of real populations.¹

He studied computer science at Swarthmore College and the University of Illinois at Urbana-Champaign.¹ His Stanford PhD, finished in 2025, was co-advised by Percy Liang and Michael Bernstein.^1,2 His dissertation, "Generative Agent Simulations of Human Behavior", won Stanford's Arthur Samuel Award, and the generative agents paper won a best paper award at UIST 2023.¹ Earlier, with the same group, he built Social Simulacra, "creating populated prototypes for social computing systems" (UIST 2022).¹ He has taught a Stanford course, CS 222, on AI agents and simulations.¹ He is co-founder and chief executive of Simile.^2,3

On the wiki he led Generative Agents and Generative Agent Simulations of 1,000 People, and Simile is his company.